One prop, two chunks of GPS mast so far!

crayfellow

Member
I have made around ten flights since this thread started and it just makes me not understand why all the hoops when the SuperX just works.
As too way-points, unless you have a commercial license when the regs come out. I can tell you the LOS restrictions are going to stick.
Now that I have 17 months and 100's of hours, there is no substitute for piloting skills even if you are flying way-points and something goes wrong. Example: say your flying way-points and your out a mile or so and a storm comes up and the winds change. Suddenly your mission is in jeopardy because it can't make it back to home against a 20 knot headwind without intervention.
I think you are simply making assumptions about others' requirements. We all do it, but it tends to lead to misunderstanding and misjudgment about others' decision making process and/or the validity of their own experiences and needs.
 

crayfellow

Member
This is what scares me, the info available is scattered and hard to find. After many flights with the SuperX I have yet to have a flyaway. And considering that Crayfellow's experience is about as close to the definition of flyaway as I would like to get, I'd say that it's safe to assume all FCs have their issues.
I think we need to be careful judging the situation until we've reviewed the data, which I've only just started doing today.

I think I would restate "all FCs have their issues" as "all FC's use essentially analogous if not identical sensor inputs and processing power, with essentially analogous if not identical algorithms. Because aggregating technology is immensely complex, this occasionally leads to confusion and/or unaccounted for edge cases on the part of the algorithms".

That is the case whether you're talking about FC's or toasters or refrigerators. The only difference is this is flying something, so you have a visible manifestation of the technology weirdness. As a dyed in the wool technologist now for 30+ years, I can tell you technology will ALWAYS exhibit weirdness whether it's flying or cooking your food or what have you. For me that's part of the fun. Perhaps for others it may not be so joyous :)
 

crayfellow

Member
Alright, here are some graphs from this experience that I think are relevant.

I was initially in Stabilize, but took off in Loiter. Once it SEEMED to be climbing inexplicably even with 0 throttle I put it back into Stabilize. This is evident in the Alt (Yellow) vs. DAlt (Desired altitude, Orange). See the altitude climb with the desired altitude, then when I hit Stabilize the motors shut off completely since the throttle was at or near 0. This is corroborated by desired altitude snapping right to 0. You can see the yellow line show the true descent, followed by me catching it a bit with manual throttle, then landing.

The question is: WHY did it think the desired altitude should be increasing when throttle input was 0?

View attachment 25609

Here you can see motor channels (Chan1, Chan2, Chan3, Chan4, Chan5, Chan6) overlaid with the C3 (Channel 3) RC input, literally my stick input. See the light purple C3 line go to 0, yet the motors were still getting power, and only cut to 0 when I went to Stabilize. In Loiter it should have started a descent at the preset rate, but instead it pushed desired altitude up.

View attachment 25610

Here is another way to look at it, ThrIn (Throttle Input) with ThrOut (Throttle Output). Throttle input goes to 0 right with my stick input, yet throttle input bounces around 50%.
View attachment 25613




Here is Mission Planner's auto analysis. I will have to check vibration (already using Kyosho Zeal pads), wire tension, and accel calibration. as for the PM "FAIL", Randy from the APM project has said "The PM failure messages is most likely a red herring. The PM messages display the slow loops after arming because arming ties up the CPU for a couple of seconds. Hopefully we will fix this false-positive for AC3.3." I will have to see if I keep getting it. Other than that this analysis seems OK, right?

View attachment 25612


Here is the accel graph in case that is useful. Z seems to be pretty crazy at the 1min mark but I have a feeling that had to do with the wind that day.

View attachment 25614


here is another look at altitude. climbing right at the 1min mark, then claiming even more when throttle went to 0. Followed by rapid descent when I switched to Stabilize, followed by my catching it at the last second, then "landing".
View attachment 25611
 

Attachments

  • Screen Shot 2015-07-18 at 4.04.02 PM.png
    Screen Shot 2015-07-18 at 4.04.02 PM.png
    119.3 KB · Views: 177
  • Screen Shot 2015-07-18 at 3.48.55 PM.png
    Screen Shot 2015-07-18 at 3.48.55 PM.png
    133.6 KB · Views: 215
  • Screen Shot 2015-07-18 at 4.01.11 PM.png
    Screen Shot 2015-07-18 at 4.01.11 PM.png
    107.7 KB · Views: 196
  • Screen Shot 2015-07-18 at 4.06.37 PM.png
    Screen Shot 2015-07-18 at 4.06.37 PM.png
    224.6 KB · Views: 195
  • Screen Shot 2015-07-18 at 4.00.42 PM.png
    Screen Shot 2015-07-18 at 4.00.42 PM.png
    86.9 KB · Views: 206
  • Screen Shot 2015-07-18 at 4.01.53 PM.png
    Screen Shot 2015-07-18 at 4.01.53 PM.png
    92.6 KB · Views: 209

Motopreserve

Drone Enthusiast
While I understand your reasoning, and appreciate your willingness to experiment and possibly lose craft and large investment - I'm not sure that I have a lack of perspective here. Thus far in my Multirotor "career" I have flown Multiwii, superx and Naza, and a few propriatary FCs. None of those provided the experience APM did.

My experience with APM 2.6 went like this....

Follow the directions I could cobble together after hopping around the Internet, dodging 404 pages and dead links on the sponsored APM site. Finally get it set as simple as humanly possible (or at least as far as I could gather), fire it up and it didn't work. After a week of trying (stressing because it was delaying a review), I slapped a Naza on the identical machine without changing a thing. Flew immediately on stock parameters.

Now I get that there are features that pixhawk has that no other FC provides. That's the appeal for sure. But if the focus can't partially include bending the technology to the code, to allow at least a basic setting to get people in the air easily - then the advanced features become even harder to obtain.

In my reading of your experience thus far - and I mean no disrespect, I see you trying to hover correctly, and Pixhawk having issues with that. This is not sending it up autonomously to chart the outback. This is hovering! :)

Of course the bottom line is its your opportunity and investment, and I would never dissuade someone from charting their own path. But it seems to back up my initial thoughts about Pixhawk making it tough at times for simply getting the thing in the air at all, so that fine tuning and advanced features can be explored.
 

crayfellow

Member
Again, I think I have simply misrepresented my intentions for this research. Also I haven't posted the numerous other autonomous flights/missions I've already done and continue to do using APM.

As for APM or Pixhawk in general I think the confusion stems from the fact that folks are judging it as a consumer ready product, and it's really not intended as such. It's the result of academic research and because of that is quite a bit more complicated than other similar devices, but has quite a bit more potential for certain applications as well.

The irony here is that this is the same kind of confusion that has given "drones" a negative connotation. These things are extremely complex. Are there some products that conceal a lot of that complexity, and work in the average case? Sure! Yet there will always be someone grabbing a Phantom or Solo off the shelf having no idea whatsoever what horizontal dilution of precision is and the thing will sail off into the sunset never to be seen again. I'm obviously not suggesting you don't know all of this.

Maybe that's the takeaway here if any. I really enjoy getting into the fine-grained detail, others would prefer (or need) those things to be concealed. A lot of times, the concealed, seemingly polished solution will be best in the average case. Except when it's not. So if you're like me and you must be prepared for any possible application, industry, or integration, you have no choice but to absolutely maximize flexibility and control and face the music to learn along the way.

When I got into cars as a kid, one of the first things I did was pull the motor and plan a completely ill-advised frankenstein of a drag racing car. Now, was the car better or worse than a sensible hatchback that is practically guaranteed to get you from A to B? Depends on your needs and interests, right?
 
Last edited by a moderator:

Motopreserve

Drone Enthusiast
I don't think you've misrepresented it - you've actually been quite clear. And I'm also not unfamiliar with the love of tinkering and being challenged by the technology.

But I would differ, to use your analogy, that the hatchback might be most sensible with a very expensive camera/gimbal on it, if it worked. And the rocket-fuel dragster might be just the ticket, when you're looking to rip around (or in a straight line in this case :) ).

The problem is there are currently no FCs that offer the features and flexibility of the Pixhawk - especially at the price point. And I get that appeal. But I wonder if the developers haven't skipped over the simple stuff in a quest for the more intriguing challenges (to them), leaving the more basic functions be either A) not as sussed out, or B) just not well documented.

In my experience, documentation was certainly an issue. I don't mind studying - but come on! Broken links to major features? And I have this issue with almost every FC manufacturer - not just the APM folks. But as I said, I've had much better luck in both flying and research with the other options.

Maybe what these projects need is a person who handles documentation. Let the geeks have their fun, but just make sure there is another dude there to document whatever the other guys are doing :)
 

crayfellow

Member
So while we likely won't solve all the world's problems in a single thread, what do you guys think about the graphs and data above? Anything stand out? I will post it on the APM forum as well in case someone can dig into the log.
 

crayfellow

Member
Posted here on the APM log review forum. I will report back if it leads to any new info.

IMO incomplete docs, etc. are typical when it comes to mixed sponsored/open source projects. You will find that with any software/hardware development environment too, even in the commercial realm. Documentation is unbelievably expensive. Maintaining documentation, even more so. The difference, I think, is the target market.

You find the same with basecam/alexmos. Most people don't know this, but "high end" gimbal manufacturers license the same technology and simply don't advertise it. Their value add is in fancy software applications and ultimately customer support, they're not green fielding new hardware technology, that would make no sense and would take years.

The same is the case with flight controllers, there is no radical difference between the supposedly "amazing" high dollar commercial ones and something like PX4 (at least, until you get to the 5-figure commercial/military level, but even then the design is essentially similar, just with more of the processing done in hardware and tighter sensor accuracy/tolerances). They are all essentially the same. The differences are in how they are pre calibrated and what simplifying assumptions the designers made. Even vibration isolation is highly dependent on the application, frame size, layout, etc. APM has to be flexible enough for tiny mini's all the way up to giant utility rigs, so it makes perfect sense to me that the "out of the box" experience may or may not appear to be as smooth as a commercial alternative. It's the difference between simplifying assumptions made in the design/tuning vs. a purposeful lack of simplifying assumptions in exchange for ultimate flexibility.

Which one is "best" for a given builder/designer/craft/etc. depends wildly on specific needs.

I'd say the same for KDE ESC's, my challenge stemmed from weak documentation on the APM wiki regarding the fixed PWM range. That is not to say KDE or Castle or any other factory calibrated opto ESC's should not be used, I think anyone who has used them will tell you they are excellent. It simply means due to the unbelievable flexibility and lack of assumptions designed into Pixhawk and APM, you need to know some things in order to use them in a design. But once those things are known, you have exactly what you need while someone with a radically different design can have exactly what they need as well.

SuperX or Naza or whatever someone else prefers cannot be guaranteed to have any better out of the box experience, you simply may happen to have a craft with properties that fall inline with the simplifying assumptions the manufacturer made.
 
Last edited by a moderator:

crayfellow

Member
Fun quick story: tricopter had a crazy crash into a very high tree branch at my folks' house last night while trying to show my father an auto mission. It ended up stuck up there, and on a canopy branch not in a place that one can climb to (at least, not me). I had disarmed and started thinking about what we could do when it started raining!

Finally I just decided to put it into Stabilize (most 'manual' APM mode), cranked the throttle, and rocked Roll back and forth until it broke free and rocketed into the sky. I somehow managed to regain control and land it only slightly roughly in the street.

Total damage: one $0.005 tie wrap, and a second to snap the collapsable GPS mast, tie wrapped landing gear, and tail servo back into place.

This morning (now that it's no longer raining) a quick flight to validate with my 'home' auto mission and everything is working great!

Perhaps it's time to start experimenting with LIDAR and optical flow for obstacle avoidance during auto missions (right @Mactadpole?)
 
Last edited by a moderator:


violetwolf

Member
Yeah, you got lucky Cray! One of mine is still in the tree 2 years later.... :/ Same scenario, rocking it with the throttle until batts ran out. Sigh
 

violetwolf

Member
PS: spot on with your observations regarding the big names simply licensing the hobby tech... I noticed that when Movi started using encoders that Basecam started offering the add on encoder firmware at almost the same time ;)

Coincidence? Lol!
 

crayfellow

Member
:D
yes indeed! Not sure why that is so offensive to people either when that is pointed out. It's sort of par for the course in product development and it wouldn't make sense at all for them to greenfield something that is well established unless they have a VERY significant design innovation. It's just silly when they come out and claim things are 100% in-house... the truth is not so thinly veiled if you know what to look for.

What upset me more is that FreeFly is "known" for the 'kinetic remote' (they call it Mimic), but BeeWorks invented it.

Bummer about yours stuck in the tree! Is it near your house so it taunts you as you come by?
 


Top